Constructing and Validating a Prognostic Model for Glycosyltransferases in Melanoma and Analyzing the Tumor Immune Microenvironment Based on Single-Cell Sequencing Data

Objective Our objective is to establish a predictive model based on glycosyltransferase-related genes (GTs) to predict the survival time of patients with Skin Cutaneous Melanoma (SKCM) and explore the pathways and mechanisms through which GTs influence the prognosis of SKCM.Methods The study utilized individualized prognostic modeling based on transcriptomic data of SKCM from the Cancer Genome Atlas (TCGA) and the reliability of the model was validated using GEO data. Univariate Cox regression and LASSO regression were employed to select biomarkers associated with SKCM prognosis and a predictive Riskscore was constructed using multivariate Cox regression. GO、KEGG and GSEA analyses were performed to annotate the functional implications of the Riskscore. The performance of the nomogram model was evaluated using ROC curves、calibration plots and C-Index. Additionally, subsequent analyses were conducted on immune infiltration、somatic mutations and im-mune therapy response based on risk stratification and scRNA analysis was employed to validate these findings.Results This study found a significant correlation between the predictive Riskscore constructed using multivariate Cox regression and the overall survival rate of SKCM. Enrichment analysis of the Riskscore revealed its association with immune function. The nomogram model which combines the Riskscore and clinical prognostic factors, demonstrated robust predictive ability in both the training and validation da-tasets. Subsequent analyses on immune infiltration、single-cell analysis、somatic mutation analysis and immune therapy response all showed a correlation between the key gene MGAT4A and the infiltration of CD8+ T cells and monocytes/macrophages in tumor tissue.Conclusion We have developed an individualized predictive model for predicting the 1-year、3-year, 5-year and 10-year survival rates of SKCM patients. This model has the potential to serve as a valuable tool in guiding personalized diagnosis and treatment for SKCM.

研究目的是基于糖基转移酶相关基因（glycosyltransferase-related genes，GTs）构建预测模型，以预测皮肤黑色素瘤（Skin Cutaneous Melanoma，SKCM）患者的生存时间，并探究GTs影响SKCM预后的通路及机制。
方法：本研究基于癌症基因组图谱（The Cancer Genome Atlas，TCGA）的SKCM转录组数据构建个体化预后模型，并通过GEO数据验证模型的可靠性。采用单变量Cox回归和LASSO回归筛选与SKCM预后相关的生物标志物，通过多变量Cox回归构建预测风险评分（Riskscore）。进行GO、KEGG及GSEA分析以注释Riskscore的功能意义。通过ROC曲线、校准图及C指数评估列线图模型的性能。此外，基于风险分层针对免疫浸润、体细胞突变及免疫治疗反应开展后续分析，并采用单细胞RNA测序分析（scRNA analysis）验证上述发现。
结果：本研究发现，通过多变量Cox回归构建的预测Riskscore与SKCM患者的总生存率显著相关。Riskscore的富集分析显示其与免疫功能相关。结合Riskscore与临床预后因素的列线图模型在训练集和验证集中均表现出稳健的预测能力。针对免疫浸润、单细胞分析、体细胞突变分析及免疫治疗反应的后续研究均显示，关键基因MGAT4A与肿瘤组织中CD8+ T细胞及单核细胞/巨噬细胞的浸润相关。
结论：本研究构建了个体化预测模型，用于预测SKCM患者1年、3年、5年及10年的生存率。该模型有望成为指导SKCM个体化诊断与治疗的重要工具。